Research And Application Of Camera Calibration Based On Data-fitting

People always have the dream that machines or robots can have the sense of vision processing ability like humankind, so machines or robots can work replace of humankind. Computer vision is the technology that solves these problems.This paper focuses on camera calibration which is an important foundation research area of computer vision. Based on camera calibration, this paper also researches measurement method based on binocular stereo vision, dynamic calibration and synthetic images.Firstly, this paper builds up the camera model which includes lens radial distortion and tangential distortion. The camera calibration procedure has two steps: first, using a direct linear transformation (DLT) to estimate the initial camera parameter values; second, solving nonlinear data-fitting to compute camera parameter values. And we also propose a sub-pixel corner detection technology based on OpenCV. A numerical implementation to correct images with pre-computed intrinsic camera parameters is presented. This camera calibration method is simple, efficient and accurate.In binocular stereo vision system, we apply the results of camera calibration for measurement and building up the epipolar-line constraint.Combining video technology and camera calibration technology, we process every frame of videos and compute the camera parameters, so we expect to dynamically calibrate cameras.Lastly, this paper introduces Augmented Reality (AR) in virtual reality environment. We describe a system for producing synthetic 3-D scenes integrated with real images captured with a camera, by using the graphic library Open Inventor. All the parts of the system are being described: real image capture, parameters measuring from the real scene, synthetic scene formation and coherent integration between the synthetic scene and the captured sequence.